A Generalizable, Data-Driven Agent-Based Transport Simulation Framework: Towards Land Use and Transport Interaction Models in Brazil

Agent-based models (ABMs) in transport represent a paradigm shift from traditional aggregate and equilibrium-based approaches. By modeling individual behaviors of a heterogeneous population, an ABM offers a more realistic representation of urban phenomena and extends sensitivity to different policy interventions. Despite this, ABM implementation faces several challenges such as limited reproducibility, uneven global implementation, and high technical and financial costs, particularly relevant in the Global South. The proposed framework addresses these gaps by implementing a modular, transparent, publicly shared data-driven approach, reducing hierarchies and relationships definitions while ensuring reproducibility. Utilizing nationally available data to generate a synthetic population, activity plans, multimodal network and agent simulations in MATSim, the framework was applied in the Metropolitan Area of Fortaleza, a region with approximately 4 million people in Brazil. Despite inherent data limitations characteristic of developing contexts, the framework demonstrated performance compatible with strategic planning applications. Traffic assignment validation showed a mean absolute error of 301 vehicles during morning peak hours and 423 vehicles for the 24 h period, which are acceptable for scenario-based policy analysis. Beyond the potential to democratize access to robust urban planning models in similar data-constrained scenarios worldwide, this study presents pathways to foster national dialogue toward improved data collection practices for disaggregated transport model implementation.